Face Feature Extraction And Recognition Based On LBP And Its Variants

Face recognition is an important research field.It has many applications in real world,such as identity verification,video surveillance etc.Many researches have done lots of work and achieved great success.In all kinds of face recognition methods,the Local Binary Pattern(LBP)for face recognition has been widely concerned for its powerful representation.In this paper,in order to solve the problems of the traditional Local binary pattern(LBP),we improve the performance with pertinence to the problems.The main contributions are as follows:1.In order to extract the features more effectively,we first conduct face location technology to locate the position of human face in the picture.Meanwhile,we also pre-process the input image by Gaussian filter to reduce interference of noise.2.In order to overcome the limitation of the traditional LBP,we divided the traditional LBP into two parts which are called CLBP_S and CLBP_M separately.The first part is CLBP_S which can encode sign information.The second part is CLBP_M which can encode magnitude information.Meanwhile,compared with to the magnitude feature,the sign feature is more powerful in texture description.Afterwards,we concatenate the histogram of both parts to get the final feature representation which is more robust.We call it CLBP_S_M which is the fusion of two complementary features.3.We choose the SVM(Support Vector Machine)as our final classifier and exploit the general method of LIBSVM(A Library for Support Vector Machines)parameters optimization.Experimental results show that our method can achieve comparable performance to the state of art.